View Metadata Map
Usage Notes
The coordinates will depend on whether the pages are automatically recognised by the OCR engine. Zoning is done by OCR, as part of the pre-processing stage. Different collections include different zoning and, as such, not all have article zones; the older collections tend to analyse data at page level. Gale segment by article, but with older newspapers do not try to figure out where one article starts and ends because it is subjective. Each article, however, has an individual title. They manually check all titles (though not the entire OCR) to make sure they are meaningful and accurate. In Delpher, article segmentation is done semi-automatically and checked by an operator. Articles are zoned rather than pages, to support users searching for specific topics and the creation of a search relevance ranking; if the word is in the article title, the ranking increases. Trove have based their decisions on other newspaper digitisation projects and what standards were used when the project began in 2006, including the Netherlands and some US projects such as the Library of Congress programme. However, they did deviate in that they OCRed and analysed the data at the article level rather than just the page level, which was common at the time.
Examples:
“In addition, during the start-up phase of a digitization project, the decision should be made if pages get rotated automatically by the OCR engine or not. If the answer is yes than the consequence is that the OCR coordinates of the rotated page either do not match the original page coordinates or that the rotated page must afterwards be regarded as a new ‘original’.” [Neudecker and Antonacopoulos, 408]
“Some form of division of the newspaper pages into ‘zones’ or ‘segments’” [Edmund King, 180]
“Optical Character Recognition works through analysis of a document image where a scanned digital image that contains machine-printed text is input into an OCR software engine and translated into a machine-readable digital text format. OCR works by first pre-processing the digital page image into its smallest component parts with layout analysis to find text blocks, sentence/line blocks, word blocks and character blocks – a process known as zoning. Other features, such as lines, graphics, photographs, etc., are recognized and may be discarded for the purpose of text recognition. The assignment of zones becomes more challenging as the layout becomes more complex, and correct zoning remains a significant issue for newspaper OCR accuracy.” [Tanner, Muñoz and Ros, par. 5]
“After raw data verification and ingest, the conversion process starts off with page analysis to determine page frames, followed by several zoning and structure recognition steps, where each element is assigned a specific zone category, and the individual elements (e.g. headlines, text blocks, illustrations) are grouped together into articles.” [Neudecker and Antonacopoulos, 408]